Know what your project can do.

Before your AI-built product reaches users, find the gaps in its code, see what was actually tested, and give your coding agent a clear next step.

Test your project Use PRAXIS locally

Try a source review in your browser. No account or project upload.

Start where your project is.

A quick browser review lets you try PRAXIS. The local app keeps your reviews beside your project and offers deeper checks when you choose.

Browser review

See the source gaps.

Select files and get findings tied to source locations, plus a prompt you can give your coding agent.

Runs in browser memory, without uploading your project
No account, model key, or Docker required
Does not run tests or prove user journeys

Try it in your browser

Local project review

Go deeper on your computer.

Review a folder, local path, or GitHub repository in a separate copy. Add models and runtime checks only when needed.

Saved findings and evidence history
Optional host or Docker execution with explicit authorization
Fix plans, agent handoff, and guarded change review

See local setupRequires Node.js 22+ and initial downloads. No Docker for the first source review.

A clear path from code to action.

Every review separates what the source suggests from what actually ran. You decide what to test or change next.

Bring your project

Choose source files in the browser, or a folder, path, or repository in the local app. Describe the behavior you expect.

Read the result

Inspect the finding, its file location, and the checks that remain unrun. Missing evidence stays visible.

Move forward

Copy a grounded agent brief or review a fix plan. Local runtime checks and edits require your authorization.

Useful answers, clear limits.

PRAXIS does not turn an AI’s “done” message into a passing test. It shows the evidence it has and the work still needed to build confidence.

Its claim-verification CLI, project memory, MCP tools, and Windows tray remain available for teams that need them. They do not replace a project review.

Explore the open-source project →

Source review: Finds code-backed concerns; it cannot prove the app works end to end.
Runtime checks: Run locally only after you choose and authorize an environment.
Model review: Optional provider calls use bounded, redacted excerpts and a shared spending limit.
Changes: Plans and prompts do not edit source. Applying proposed changes is a separate decision.

Questions before you start.

Will my project be uploaded?

The browser trial reads selected files in browser memory and makes no project upload. Local reviews keep a separate copy on your computer. GitHub intake, optional advisory checks, and connected model calls use the network when selected.

Do I need Docker or an API key?

No. The first source review needs neither. Docker or trusted host execution is optional for runtime checks; model keys are only needed for optional model review or AI-assisted planning.

Can PRAXIS say my app is production-ready?

It reports what its evidence supports. Source inspection is a first check, not proof of user-facing behavior or production operations. Deeper claims need relevant tests, observed journeys, and deployment evidence.

Will PRAXIS fix my code automatically?

No. It can prepare an agent brief or a fix plan. You review the handoff and separately authorize any edit or application of a proposed change.

Find the next step before your users find the gap.

Start with a browser source review, then move locally when you need more evidence.

Test your project Use PRAXIS locally